Message ID | 20230912-topic-tps6598x_reset-v1-1-78dc0bf61790@wolfvision.net (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Series | usb: typec: tps6598x: add reset gpio support | expand |
On 15/09/2023 08:50, Javier Carrasco wrote: > The TPS6598x PD controller provides an active-high hardware reset input > that reinitializes all device settings. If it is not grounded by > design, the driver must be able to de-assert it in order to initialize > the device. > > The PD controller is not ready for registration right after the reset > de-assertion and a delay must be introduced in that case. According to > TI, the delay can reach up to 1000 ms [1], which is in line with the > experimental results obtained with a TPS65987D. > > Add a GPIO descriptor for the reset signal and basic reset management > for initialization and suspend/resume. > > [1] https://e2e.ti.com/support/power-management-group/power-management/ > f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- > to-normal-operation/4809389#4809389 > > Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> > --- > drivers/usb/typec/tipd/core.c | 21 ++++++++++++++++++++- > 1 file changed, 20 insertions(+), 1 deletion(-) > > diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c > index 37b56ce75f39..550f5913e985 100644 > --- a/drivers/usb/typec/tipd/core.c > +++ b/drivers/usb/typec/tipd/core.c > @@ -8,6 +8,7 @@ > > #include <linux/i2c.h> > #include <linux/acpi.h> > +#include <linux/gpio/consumer.h> > #include <linux/module.h> > #include <linux/of.h> > #include <linux/power_supply.h> > @@ -43,6 +44,9 @@ > /* TPS_REG_SYSTEM_CONF bits */ > #define TPS_SYSCONF_PORTINFO(c) ((c) & 7) > > +/* reset de-assertion to ready for operation */ > +#define SETUP_MS 1000 1 second? That's a bit a lot. Does it come from datasheet? > + > enum { > TPS_PORTINFO_SINK, > TPS_PORTINFO_SINK_ACCESSORY, > @@ -86,6 +90,7 @@ struct tps6598x { > struct mutex lock; /* device lock */ > u8 i2c_protocol:1; > > + struct gpio_desc *reset; > struct typec_port *port; > struct typec_partner *partner; > struct usb_pd_identity partner_identity; > @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) > mutex_init(&tps->lock); > tps->dev = &client->dev; > > + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", GPIOD_OUT_LOW); > + if (IS_ERR(tps->reset)) > + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), > + "failed to get reset GPIO\n"); > + if (tps->reset) > + msleep(SETUP_MS); > + > tps->regmap = devm_regmap_init_i2c(client, &tps6598x_regmap_config); > if (IS_ERR(tps->regmap)) > return PTR_ERR(tps->regmap); > @@ -892,6 +904,9 @@ static void tps6598x_remove(struct i2c_client *client) > tps6598x_disconnect(tps, 0); > typec_unregister_port(tps->port); > usb_role_switch_put(tps->role_sw); > + > + if (tps->reset) > + gpiod_set_value_cansleep(tps->reset, 1); > } > > static int __maybe_unused tps6598x_suspend(struct device *dev) > @@ -902,7 +917,8 @@ static int __maybe_unused tps6598x_suspend(struct device *dev) > if (tps->wakeup) { > disable_irq(client->irq); > enable_irq_wake(client->irq); > - } > + } else if (tps->reset) > + gpiod_set_value_cansleep(tps->reset, 1); Missing {} (see Linux coding style). > > if (!client->irq) > cancel_delayed_work_sync(&tps->wq_poll); > @@ -918,6 +934,9 @@ static int __maybe_unused tps6598x_resume(struct device *dev) > if (tps->wakeup) { > disable_irq_wake(client->irq); > enable_irq(client->irq); > + } else if (tps->reset) { > + gpiod_set_value_cansleep(tps->reset, 0); > + msleep(SETUP_MS); > } > > if (!client->irq) > Best regards, Krzysztof
Hello Krzysztof, On 15.09.23 08:57, Krzysztof Kozlowski wrote: > On 15/09/2023 08:50, Javier Carrasco wrote: >> The TPS6598x PD controller provides an active-high hardware reset input >> that reinitializes all device settings. If it is not grounded by >> design, the driver must be able to de-assert it in order to initialize >> the device. >> >> The PD controller is not ready for registration right after the reset >> de-assertion and a delay must be introduced in that case. According to >> TI, the delay can reach up to 1000 ms [1], which is in line with the >> experimental results obtained with a TPS65987D. >> >> Add a GPIO descriptor for the reset signal and basic reset management >> for initialization and suspend/resume. >> >> [1] https://e2e.ti.com/support/power-management-group/power-management/ >> f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- >> to-normal-operation/4809389#4809389 >> >> Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> >> --- >> drivers/usb/typec/tipd/core.c | 21 ++++++++++++++++++++- >> 1 file changed, 20 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c >> index 37b56ce75f39..550f5913e985 100644 >> --- a/drivers/usb/typec/tipd/core.c >> +++ b/drivers/usb/typec/tipd/core.c >> @@ -8,6 +8,7 @@ >> >> #include <linux/i2c.h> >> #include <linux/acpi.h> >> +#include <linux/gpio/consumer.h> >> #include <linux/module.h> >> #include <linux/of.h> >> #include <linux/power_supply.h> >> @@ -43,6 +44,9 @@ >> /* TPS_REG_SYSTEM_CONF bits */ >> #define TPS_SYSCONF_PORTINFO(c) ((c) & 7) >> >> +/* reset de-assertion to ready for operation */ >> +#define SETUP_MS 1000 > > 1 second? That's a bit a lot. Does it come from datasheet? > I also thought it is a long delay, but when I tested it myself with real hardware I got values of hundreds of milliseconds until the regmap_init_i2c was able to reach the device. I also noticed that there is already a timeout of 1s in the tps_6598x_exec_cmd function, which made me suspect that the timing might be a factor to consider. The datasheet does not provide any timing for the reset, so I asked the manufacturer (TI). The reply form the TI technical staff was that depending on the configuration it might take from 800 to 1000 ms. I added a link to that answer in the commit message. >> + >> enum { >> TPS_PORTINFO_SINK, >> TPS_PORTINFO_SINK_ACCESSORY, >> @@ -86,6 +90,7 @@ struct tps6598x { >> struct mutex lock; /* device lock */ >> u8 i2c_protocol:1; >> >> + struct gpio_desc *reset; >> struct typec_port *port; >> struct typec_partner *partner; >> struct usb_pd_identity partner_identity; >> @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) >> mutex_init(&tps->lock); >> tps->dev = &client->dev; >> >> + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", GPIOD_OUT_LOW); >> + if (IS_ERR(tps->reset)) >> + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), >> + "failed to get reset GPIO\n"); >> + if (tps->reset) >> + msleep(SETUP_MS); >> + >> tps->regmap = devm_regmap_init_i2c(client, &tps6598x_regmap_config); >> if (IS_ERR(tps->regmap)) >> return PTR_ERR(tps->regmap); >> @@ -892,6 +904,9 @@ static void tps6598x_remove(struct i2c_client *client) >> tps6598x_disconnect(tps, 0); >> typec_unregister_port(tps->port); >> usb_role_switch_put(tps->role_sw); >> + >> + if (tps->reset) >> + gpiod_set_value_cansleep(tps->reset, 1); >> } >> >> static int __maybe_unused tps6598x_suspend(struct device *dev) >> @@ -902,7 +917,8 @@ static int __maybe_unused tps6598x_suspend(struct device *dev) >> if (tps->wakeup) { >> disable_irq(client->irq); >> enable_irq_wake(client->irq); >> - } >> + } else if (tps->reset) >> + gpiod_set_value_cansleep(tps->reset, 1); > > Missing {} (see Linux coding style). > Thanks, I will correct this for v2. >> >> if (!client->irq) >> cancel_delayed_work_sync(&tps->wq_poll); >> @@ -918,6 +934,9 @@ static int __maybe_unused tps6598x_resume(struct device *dev) >> if (tps->wakeup) { >> disable_irq_wake(client->irq); >> enable_irq(client->irq); >> + } else if (tps->reset) { >> + gpiod_set_value_cansleep(tps->reset, 0); >> + msleep(SETUP_MS); >> } >> >> if (!client->irq) >> > > Best regards, > Krzysztof > Best regards, Javier Carrasco
On 15/09/2023 07:50, Javier Carrasco wrote: > The TPS6598x PD controller provides an active-high hardware reset input > that reinitializes all device settings. If it is not grounded by > design, the driver must be able to de-assert it in order to initialize > the device. > > The PD controller is not ready for registration right after the reset > de-assertion and a delay must be introduced in that case. According to > TI, the delay can reach up to 1000 ms [1], which is in line with the > experimental results obtained with a TPS65987D. > > Add a GPIO descriptor for the reset signal and basic reset management > for initialization and suspend/resume. > > [1] https://e2e.ti.com/support/power-management-group/power-management/ > f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- > to-normal-operation/4809389#4809389 > > Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> > --- > drivers/usb/typec/tipd/core.c | 21 ++++++++++++++++++++- > 1 file changed, 20 insertions(+), 1 deletion(-) > > diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c > index 37b56ce75f39..550f5913e985 100644 > --- a/drivers/usb/typec/tipd/core.c > +++ b/drivers/usb/typec/tipd/core.c > @@ -8,6 +8,7 @@ > > #include <linux/i2c.h> > #include <linux/acpi.h> > +#include <linux/gpio/consumer.h> > #include <linux/module.h> > #include <linux/of.h> > #include <linux/power_supply.h> > @@ -43,6 +44,9 @@ > /* TPS_REG_SYSTEM_CONF bits */ > #define TPS_SYSCONF_PORTINFO(c) ((c) & 7) > > +/* reset de-assertion to ready for operation */ > +#define SETUP_MS 1000 > + > enum { > TPS_PORTINFO_SINK, > TPS_PORTINFO_SINK_ACCESSORY, > @@ -86,6 +90,7 @@ struct tps6598x { > struct mutex lock; /* device lock */ > u8 i2c_protocol:1; > > + struct gpio_desc *reset; > struct typec_port *port; > struct typec_partner *partner; > struct usb_pd_identity partner_identity; > @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) > mutex_init(&tps->lock); > tps->dev = &client->dev; > > + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", GPIOD_OUT_LOW); > + if (IS_ERR(tps->reset)) > + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), > + "failed to get reset GPIO\n"); > + if (tps->reset) > + msleep(SETUP_MS); > + This looks a bit odd to me, shouldn't you drive reset to zero ? if (tps->reset) { gpiod_set_value_cansleep(tps->reset, 0); msleep(SETUP_MS); } also wouldn't it make sense to functionally decompose that and reuse in probe() and tps6598x_resume() ? tps6598x_reset() { if (tps->reset) { gpiod_set_value_cansleep(tps->reset, 0); msleep(SETUP_MS); } } --- bod
On 15/09/2023 11:52, Bryan O'Donoghue wrote: >> + struct gpio_desc *reset; >> struct typec_port *port; >> struct typec_partner *partner; >> struct usb_pd_identity partner_identity; >> @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) >> mutex_init(&tps->lock); >> tps->dev = &client->dev; >> >> + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", GPIOD_OUT_LOW); >> + if (IS_ERR(tps->reset)) >> + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), >> + "failed to get reset GPIO\n"); >> + if (tps->reset) >> + msleep(SETUP_MS); >> + > > This looks a bit odd to me, shouldn't you drive reset to zero ? > > if (tps->reset) { > gpiod_set_value_cansleep(tps->reset, 0); It is driven by GPIOD_OUT_LOW. > msleep(SETUP_MS); > } > > also wouldn't it make sense to functionally decompose that and reuse in > probe() and tps6598x_resume() ? > > tps6598x_reset() { > if (tps->reset) { > gpiod_set_value_cansleep(tps->reset, 0); > msleep(SETUP_MS); > } > } Best regards, Krzysztof
Hello Bryan, On 15.09.23 11:52, Bryan O'Donoghue wrote: > On 15/09/2023 07:50, Javier Carrasco wrote: >> The TPS6598x PD controller provides an active-high hardware reset input >> that reinitializes all device settings. If it is not grounded by >> design, the driver must be able to de-assert it in order to initialize >> the device. >> >> The PD controller is not ready for registration right after the reset >> de-assertion and a delay must be introduced in that case. According to >> TI, the delay can reach up to 1000 ms [1], which is in line with the >> experimental results obtained with a TPS65987D. >> >> Add a GPIO descriptor for the reset signal and basic reset management >> for initialization and suspend/resume. >> >> [1] >> https://e2e.ti.com/support/power-management-group/power-management/ >> f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- >> to-normal-operation/4809389#4809389 >> >> Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> >> --- >> drivers/usb/typec/tipd/core.c | 21 ++++++++++++++++++++- >> 1 file changed, 20 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/usb/typec/tipd/core.c >> b/drivers/usb/typec/tipd/core.c >> index 37b56ce75f39..550f5913e985 100644 >> --- a/drivers/usb/typec/tipd/core.c >> +++ b/drivers/usb/typec/tipd/core.c >> @@ -8,6 +8,7 @@ >> #include <linux/i2c.h> >> #include <linux/acpi.h> >> +#include <linux/gpio/consumer.h> >> #include <linux/module.h> >> #include <linux/of.h> >> #include <linux/power_supply.h> >> @@ -43,6 +44,9 @@ >> /* TPS_REG_SYSTEM_CONF bits */ >> #define TPS_SYSCONF_PORTINFO(c) ((c) & 7) >> +/* reset de-assertion to ready for operation */ >> +#define SETUP_MS 1000 >> + >> enum { >> TPS_PORTINFO_SINK, >> TPS_PORTINFO_SINK_ACCESSORY, >> @@ -86,6 +90,7 @@ struct tps6598x { >> struct mutex lock; /* device lock */ >> u8 i2c_protocol:1; >> + struct gpio_desc *reset; >> struct typec_port *port; >> struct typec_partner *partner; >> struct usb_pd_identity partner_identity; >> @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) >> mutex_init(&tps->lock); >> tps->dev = &client->dev; >> + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", >> GPIOD_OUT_LOW); >> + if (IS_ERR(tps->reset)) >> + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), >> + "failed to get reset GPIO\n"); >> + if (tps->reset) >> + msleep(SETUP_MS); >> + > > This looks a bit odd to me, shouldn't you drive reset to zero ? > > if (tps->reset) { > gpiod_set_value_cansleep(tps->reset, 0); > msleep(SETUP_MS); > } > The reset line is driven to zero by means fo the GPIOD_OUT_LOW flag, so there is no need to set it explicitly again. > also wouldn't it make sense to functionally decompose that and reuse in > probe() and tps6598x_resume() ? > > tps6598x_reset() { > if (tps->reset) { > gpiod_set_value_cansleep(tps->reset, 0); > msleep(SETUP_MS); > } > } > I can move the reset action to a separate function as you proposed, but then I suppose it would make sense to use the same function for both reset levels. Maybe something like: tps6598x_reset(bool level) { if (tps->reset) { gpiod_set_value_cansleep(tps->reset, level); if (!level) msleep(SETUP_MS); } } > --- > bod Thanks for your feedback. Best regards, Javier Carrasco
On 15/09/2023 07:50, Javier Carrasco wrote: > The TPS6598x PD controller provides an active-high hardware reset input > that reinitializes all device settings. If it is not grounded by > design, the driver must be able to de-assert it in order to initialize > the device. > > The PD controller is not ready for registration right after the reset > de-assertion and a delay must be introduced in that case. According to > TI, the delay can reach up to 1000 ms [1], which is in line with the > experimental results obtained with a TPS65987D. > > Add a GPIO descriptor for the reset signal and basic reset management > for initialization and suspend/resume. > > [1] https://e2e.ti.com/support/power-management-group/power-management/ > f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- > to-normal-operation/4809389#4809389 > > Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
diff --git a/drivers/usb/typec/tipd/core.c b/drivers/usb/typec/tipd/core.c index 37b56ce75f39..550f5913e985 100644 --- a/drivers/usb/typec/tipd/core.c +++ b/drivers/usb/typec/tipd/core.c @@ -8,6 +8,7 @@ #include <linux/i2c.h> #include <linux/acpi.h> +#include <linux/gpio/consumer.h> #include <linux/module.h> #include <linux/of.h> #include <linux/power_supply.h> @@ -43,6 +44,9 @@ /* TPS_REG_SYSTEM_CONF bits */ #define TPS_SYSCONF_PORTINFO(c) ((c) & 7) +/* reset de-assertion to ready for operation */ +#define SETUP_MS 1000 + enum { TPS_PORTINFO_SINK, TPS_PORTINFO_SINK_ACCESSORY, @@ -86,6 +90,7 @@ struct tps6598x { struct mutex lock; /* device lock */ u8 i2c_protocol:1; + struct gpio_desc *reset; struct typec_port *port; struct typec_partner *partner; struct usb_pd_identity partner_identity; @@ -717,6 +722,13 @@ static int tps6598x_probe(struct i2c_client *client) mutex_init(&tps->lock); tps->dev = &client->dev; + tps->reset = devm_gpiod_get_optional(tps->dev, "reset", GPIOD_OUT_LOW); + if (IS_ERR(tps->reset)) + return dev_err_probe(tps->dev, PTR_ERR(tps->reset), + "failed to get reset GPIO\n"); + if (tps->reset) + msleep(SETUP_MS); + tps->regmap = devm_regmap_init_i2c(client, &tps6598x_regmap_config); if (IS_ERR(tps->regmap)) return PTR_ERR(tps->regmap); @@ -892,6 +904,9 @@ static void tps6598x_remove(struct i2c_client *client) tps6598x_disconnect(tps, 0); typec_unregister_port(tps->port); usb_role_switch_put(tps->role_sw); + + if (tps->reset) + gpiod_set_value_cansleep(tps->reset, 1); } static int __maybe_unused tps6598x_suspend(struct device *dev) @@ -902,7 +917,8 @@ static int __maybe_unused tps6598x_suspend(struct device *dev) if (tps->wakeup) { disable_irq(client->irq); enable_irq_wake(client->irq); - } + } else if (tps->reset) + gpiod_set_value_cansleep(tps->reset, 1); if (!client->irq) cancel_delayed_work_sync(&tps->wq_poll); @@ -918,6 +934,9 @@ static int __maybe_unused tps6598x_resume(struct device *dev) if (tps->wakeup) { disable_irq_wake(client->irq); enable_irq(client->irq); + } else if (tps->reset) { + gpiod_set_value_cansleep(tps->reset, 0); + msleep(SETUP_MS); } if (!client->irq)
The TPS6598x PD controller provides an active-high hardware reset input that reinitializes all device settings. If it is not grounded by design, the driver must be able to de-assert it in order to initialize the device. The PD controller is not ready for registration right after the reset de-assertion and a delay must be introduced in that case. According to TI, the delay can reach up to 1000 ms [1], which is in line with the experimental results obtained with a TPS65987D. Add a GPIO descriptor for the reset signal and basic reset management for initialization and suspend/resume. [1] https://e2e.ti.com/support/power-management-group/power-management/ f/power-management-forum/1269856/tps65987d-tps65987d-reset-de-assert- to-normal-operation/4809389#4809389 Signed-off-by: Javier Carrasco <javier.carrasco@wolfvision.net> --- drivers/usb/typec/tipd/core.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)